Aline Fuchs, Mickael Audrain, Romain Ollier, Florence Guilhot, Madiha Derouazi, Andrea Pfeifer, Marie Kosco-Vilbois, Tamara Seredenina, Tariq Afroz
AC Immune SA, Lausanne, Switzerland
Introduction: Providing a prediction of the efficacious dose needed to treat patients is a key element to support the developability of a monoclonal antibody (mAb) and design of the initial clinical trial. This case study presents our approach using target engagement characterization of a mAb for central nervous system (CNS) indication that binds to extracellular protein target.
Objective: Leverage the preclinical data from human in vitro models and non-human primate (NHP) in vivo target occupancy studies to support clinical development by providing human dose prediction of a mAb whose target is within the CNS.
Methods: For the human PK prediction of mAb, ACI-2, serum mAb concentrations were determined by ELISA after a single intravenous (iv) bolus administration of 40 mg/kg to NHP (n=3) with sampling over 8 weeks. A 2‑compartmental model with linear elimination (non-linear mixed effect model NLME approach, Monolix 2019R1) best characterized the PK profile. Typical PK parameters were scaled by allometry to predict 70 kg human PK parameters. To establish the PK/pharmacodynamic (PD) relationship for ACI-2, an in vitro target engagement assay was established to determine the concentration of ACI-2 required for target saturation in human CSF (n=3 donors). Total and free target levels in sera were also measured over 8 weeks in a PK study in NHPs for a second mAb, ACI-1 (that has the same target binding affinity as ACI‑2), tested at 4 mg/kg and 40 mg/kg iv (n=4 NHP in each dosing group). Using these data, a serum exposure‑target occupancy relationship was established via an Emax model (Rstudio 3.6.3, nls R package). For the human dose prediction of ACI-2, human PK profile for various doses was simulated based on point-estimate PK parameters and simulating a virtual population (n=200) with 20% coefficient variation (CV) attributed on each PK parameter to account for inter-individual variability. Simulations of human predicted concentration-time profile were performed to support a 4-weekly iv administration (IQRtools 1.16.1) of ACI-2. The human efficacious dose was defined as the dose needed to maintain the cerebrospinal fluid (CSF) concentration above the target binding of 90% (EC90) during all dosing interval either based on point-estimate or for 90% of the simulated population. The standard assumption of 0.1% serum to CSF penetration was used for systemically administered mAbs.
Results: Based on the prolonged exposure data of ACI-2 observed in NHP, low clearance of <0.05 mL/h/kg and volume of distribution (i.e., 60 mL/kg) were predicted for humans. Target engagement studies led to similar results reinforcing confidence in the target occupancy quantification. From the in vitro target engagement assay of ACI-2 in human CSF, an EC50 of 130 ng/mL and an EC90 of 1100 ng/mL were determined. From the in vivo PK study in NHPs with ACI-1, an EC50 of 200 ng/mL was determined in serum. Administering 40 mg/kg iv of ACI-2 in the NHP PK study, a maximum effect to achieve full elimination of the free target in serum was observed at a concentration of >1000 ng/mL; further EC50 determination was not established in absence of suboptimal exposure. The in vitro human CSF EC90 was used as target value for the prediction of efficacious dose in human. Simulations carried out based on point-estimate PK parameters resulted in a human dose prediction of approximately 50 mg/kg. The addition of 20% interindividual variability on PK parameters led to a predicted dose of approximately 120 mg/kg to 90% of the simulated population.
Conclusions: Obtaining both in vitro and in vivo target occupancy data as well as PK profile of the mAb post injection from sera of the preclinical species provided key parameter estimates to determine a provisional predicted human dose required to be efficacious in clinical trials for CNS indication. This result can be used to inform rational dose regimens in future clinical studies. This prediction will be further refined with toxicokinetic NHP data in the human PK prediction and any other relevant PK/PD data.
Reference: PAGE 32 (2024) Abstr 10843 [www.page-meeting.org/?abstract=10843]
Poster: Drug/Disease Modelling - CNS